in the wake of the 2008/09 financial crisis, the...
TRANSCRIPT
1
FLUCTUATIONS IN GOVERNMENT SIZE IN THE OECD: 1973-2011
James Obben1
School of Economics and Finance
Massey University
New Zealand
ABSTRACT
In the wake of the 2008/09 financial crisis, the concurrence of a ‘large’ government size (GS)
and slow economic growth has become quite stark in many countries. The widespread
expectation for governments to scale back assumes public sector downsizing is uniformly
growth-enhancing for all countries. Empirical evidence, however, is mixed. The study revisits
the GS-growth issue by analysing a balanced panel dataset for 24 OECD countries that
includes the most recent data. It investigates four aspects of the issue where past studies have
considered only one or two: the impact of GS on growth; the existence of the inverted-U
shaped relation between GS and economic growth (the Barro-Armey-Rahn-Scully or BARS
curve) and the resultant optimal GS; the differential impacts of GS depending on the growth
rate using quantile regression; decomposition of the GS time series to determine whether the
shocks to it leave permanent or transitory effects on its level. GS here is government final
consumption expressed as a percentage of GDP. The linear model shows a significant
negative effect of GS on growth. The BARS curve was confirmed for 13 out of the 24
countries, and among these the period average of GS exceeds the optimal size. For those
countries, downsizing could be growth-enhancing but not necessarily so for the other
countries. The quantile regression results show that the impact of GS is positive and
insignificant at low rates of economic growth and continually decreases as growth increases
until at some rate it turns negative and progressively significant with the growth rate. The
Hodrick-Prescott decomposition indicates that in all countries, except Australia, the shocks to
GS leave predominantly permanent (rather than transitory) effects.
Key words: government size, quantile regression, Hodrick-Prescott technique, OECD.
JEL: E62, H5, O4, O5.
1. INTRODUCTION
World Bank data indicate that government final consumption expenditures as a percentage of
GDP (henceforth, government size) for the whole world fluctuated upward from 14% in
1961, peaked at 18.9% in 2009 and fell slightly to 18.5% in 2011. Correspondingly, the world
economic growth rate of 4.4% in 1961 fluctuated thereafter reaching a nadir of -2.2% in 2009
and rose to 2.7% in 2011. The correlation between government size and economic growth
rate over those years is -0.67 for the world and -0.72 for the OECD. In each of the 51 years,
the average government size in the OECD countries, the group of countries which accounts
for an average of 82% of world GDP annually, is bigger than the world average government
size and the gap has been widening steadily over the last eight years. And especially in the
1 Email: [email protected]. Tel.: (+64) 6 356 9099. Fax: (+64) 6 350 5660.
2
wake of the 2008/09 global financial crisis, the concurrence of a ‘large’ government size and
slow economic growth has become quite stark in many countries. The widespread
expectation for governments to scale back assumes public sector downsizing is uniformly
growth-enhancing for all countries. Empirical evidence, however, is mixed although an
overwhelming majority of the studies on government size and growth surveyed by Bergh and
Henrekson (2011) found significant negative growth effect from government size. The
problems of large government size are highlighted by the current onerously high debt levels,
slow growth and high unemployment observed in the US and the euro-zone.
Given these observations and the primacy of the OECD in the world economy, the study
revisits the government size-growth issue by analysing a balanced panel dataset for 24 OECD
countries that includes the most recent data. It investigates four aspects of the issue where
past studies have considered only one or two: the impact of government size on growth; the
existence of the hypothesised inverted-U shaped relation between government size and
economic growth (the Barro-Armey-Rahn-Scully or BARS curve) and the resultant growth-
maximising or ‘optimal’ government size; the differential impacts of government size
depending on the growth rate using quantile regression; decomposition of the government
size time series to determine whether the shocks to it leave permanent or transitory effects on
its level.
The twenty countries that founded the OECD in 1961 have, over time, been joined by
fourteen others as at the end of 2012. Available World Bank data on these countries start in
1961 and end in 2011 at the time of this study. To obtain the longest time series for the
largest number of countries based on the years of accession of member countries, the decision
was made to include the countries which have been members since 1973. That approach
yielded a sample of 24 countries with 39 years of data. The names and international iso-codes
of the countries are: Australia (AUS), Austria (AUT), Belgium (BEL), Canada (CAN),
Denmark (DNK), Finland (FIN), France (FRA), Germany (GER), Greece (GRC), Iceland
(ISL), Ireland (IRL), Italy (ITA), Japan (JPN), Luxembourg (LUX), Netherlands (NLD),
New Zealand (NZL), Norway (NOR), Portugal (PRT), Spain (ESP), Sweden (SWE),
Switzerland (CHE), Turkey (TUR), United Kingdom (GBR) and United States (USA).
Tests of the impact of government size on economic growth with panel regression models
suggested a strong negative impact and, although the BARS curve was confirmed for the
whole group, the evidence was weak. The group growth-maximising government size was
estimated to be 7.4%, much lower than the group period average of government size of 19%.
Individual country BARS curves were also estimated and the diversity noted. The quantile
regression models for the whole sample suggest that the impact of government size is positive
at low rates of growth, declines as the growth rate increases and eventually turns negative.
Employing the Hodrick-Prescott filtering technique, the government size time series of each
country was decomposed into its trend/permanent and cycle/transitory components. From the
estimated variances of the components, it was inferred that the changes in government size
for virtually all the countries are predominantly of the permanent nature. The rest of the paper
is structured as follows. Section 2 reviews the literature on government size and its effects on
the economy and the literature on the decomposition of time series. Section 3 deals with the
data and methodology. Section 4 presents and discusses the analytical results and Section 5
concludes the paper.
3
2. LITERATURE
2.1 The Growth-Government Size Relationship
The main economic reason for the continuing concern with government size is the
preponderance of the negative correlation with economic growth found in empirical studies
(Bergh and Henrekson, 2011). Economic theory, however, is uncertain about the direction of
the impact of government size on the rate of economic growth; persuasive arguments can be
made for both positive and negative impacts of government size. Traditionally, one point of
view asserts that a larger government size can be a powerful engine for growth through the
development of a legal, administrative and economic infrastructure and by securing an
increase in productive investment and providing a socially optimal direction for growth and
development. The government also plays the crucial role of harmonising conflicts between
private and social interests and the prevention of exploitation of the country by foreigners.
This point of view may be said to use the theory of market failures to justify state
interventionism. An opposing point of view holds that as the public sector increases in size,
greater and greater taxation needs to be extracted to finance the expenditure. The increasing
taxation distorts economic incentives to work, save and invest and lowers the productivity of
the system. The economy slows down as a result. This point of view may be said to use the
theory of government failures to vilify large government size.
The relationship between government size and economic growth was at first studied in the
framework of a linear model of a Cobb-Douglas production function pioneered by Feder
(1982) but has since been investigated with a variety of analytical techniques. Mirroring the
different points of view, some of the studies have found positive correlation between
government size and economic growth (e.g., Ram, 1986; Kormendi and Meguire, 1986;
Grossman, 1990; Dar and AmirKhalkhali, 2002) whilst others have found negative
relationship between government size and economic growth (e.g., Cameron, 1982; Landau,
1985; Saunders, 1985, 1986; Guseh, 1997; Gwartney et al., 1998; Tanninen, 1999; Fölster
and Henrekson, 2001; Dar and AmirKhalkhali, 2002; Bergh and Henrekson, 2011; and
Afonso and Jalles, 2011). Unsurprisingly, other studies have reported mixed or insignificant
results (e.g., Conte and Darrat, 1988; Ghali, 1998; Grimes, 2003). Most empirical studies
have used the linear model; other studies, proferring that the conflicting results could be due
to a nonlinear relationship have investigated and found nonlinear impact of government size
on economic growth (e.g., Barro, 1990; Sheehey, 1993; Armey, 1995; Rahn and Fox, 1996;
Vedder and Galloway, 1998; Scully, 1998, 2003; and Chen and Lee, 2005). The underlying
argument is that a modicum of government is required to avoid anarchy and to foster
economic activity and growth up to a certain extent; beyond that a bloated government can
impede growth through inefficiency and diminishing marginal productivity. The resulting
inverted-U shape of the relationship is named the Barro-Armey-Rahn-Scully or BARS curve2
after these researchers who popularised the concept and made it a working tool of
contemporary economic analysis (Chobanov and Mladenova, 2009; Forte and Magazzino,
2010; Facchini and Melki, 2011). The peak of the BARS curve identifies the growth-
maximising size of government or the optimum government size. On why there have been
conflicting empirical results, the New Zealand Treasury (2011, pp. 9-10) notes that,
2 The ‘BARS curve’ is similar to the ‘Laffer curve’ (Laffer, 2004) in shape except that the latter shows the
relationship between tax revenue and tax rates and the former shows the relationship between economic growth
rate on the vertical axis and government size on the horizontal axis. The BARS curve has variously been
referred to as the Armey curve, the Rahn curve and the Scully curve.
4
‘Though some types of expenditure enhance economic growth, at some point the
economic costs of raising taxes to fund that expenditure will outweigh its benefits.
This suggests that there is an optimal level of government expenditure, from an
economic growth perspective, which balances the economic benefits of expenditure
against the economic costs of taxes. However, the economic growth impacts will
depend on the tax and expenditure mix as well as the level of expenditure.’
The empirical studies on the government size-growth relationship usually estimate cross-
country growth regressions that attempt to take account of the range of determinants of
economic growth. The determinants include ‘state’ variables (physical and human capital and
labour) and ‘control’ variables such as the ratio of government consumption or total
expenditure to GDP, initial GDP, share of domestic investment in GDP, trade openness,
movements in the terms of trade, indicators of macroeconomic stability, the fertility rate and
institutional quality (Barro and Sala-i-Martin, 2003; Afonso and Jalles, 2011). From the
simple bivariate correlation between government size and growth reported by the pioneers
(e.g., Smith, 1975; Cameron, 1982; Saunders, 1985) the analytical techniques employed have
progressed through multivariate pooled cross-section OLS regression models (Barro 1991;
Barro and Sala-i-Martin, 1992), cointegration and vector error correction models (Ghali,
1998), panel fixed and random effects estimations (e.g., Fölster and Henrekson, 2001;
Romero-Avila and Strauch, 2008), random coefficients model (e.g., Dar and AmirKhalkhali,
2002), threshold regression methodology (Chen and Lee, 2005), smooth transition
autoregressive (STAR) framework (Chiou-Wei et al., 2010); quantile regression (e.g., Chen
et al., 2011), and polynomial analysis (Herath, 2012). The variation in the results could be
attributed to the differences in analytical methods, country groups, control variables,
indicators for government size and economic growth,3 data types and sample periods that
have been used. Other nontrivial methodological problems (such as endogeneity and data
issues) trouble all the studies.
A number of critical reviews of the empirical evidence have reported that alternative proxies
of government size have produced contradictory or insignificant results (Agell et al., 1997;
Temple, 1999; Myles, 2000; and Nijkamp and Poot, 2004). The summary of the empirical
evidence is perhaps best captured with the characterisation offered by Barro and Sala-i-
Martin (2003) that government size did have negative impact on economic growth in simple
models with a relatively small number of variables; however, it became insignificant when
additional variables were added to the model, suggesting that government size was capturing
the impact of these other variables in the simple models. The generality of this caveat is
exemplified by the finding by Afonso and Jalles that, in their relatively sophisticated panel
methodology, when the European Commission numerical fiscal rules are controlled for, the
previously very significant negative impact of government size for European Union member
countries fizzles into insignificance (2011, p. 20). In view of all of this, the New Zealand
Treasury 2025 Taskforce (2010) recommends that the results of cross-country regressions
should be read as illustrative rather than determinative.
3 For economic growth some studies used the growth rate of real GDP (e.g., Dar and AmarKhalkali, 2002; Chen
et al., 2011; Herath, 2012) and others used the growth rate of real GDP per capita (e.g., Chiou-Wei et al., 2010).
Proxies of government size that have been used include government total expenditure as a ratio of GDP (e.g.,
Dar and AmarKhalkali, 2002; Chen et al., 2011; Herath, 2012), government final consumption as a ratio of GDP
(e.g., Chiou-Wei et al., 2010), tax revenue as a ratio of GDP (e.g., Agell et al., 2006; Bergh and Henrekson,
2011).
5
As noted earlier, in most of the studies, linear regression was used. Linear regression is a
conditional mean approach which means the estimated coefficients indicate how much the
expected value of the dependent variable (given the values of the independent variables) is
expected to change given a one unit change in the variables the coefficients represent, ceteris
paribus. In the growth models the coefficient of the government size variable reflects the
impact of government size on the growth rate of countries with growth rates similar to the
sample mean. The coefficient cannot reflect the growth impacts on countries with growth
rates lower or higher than the average growth rate. Those effects can be captured by using
quantile regression. Quantile regression shows the relation between a set of independent
variables and specific percentiles (quantiles) of the distribution of the dependent variable.
Therefore, a growth quantile regression parameter for government size estimates the growth
impact for a specified quantile of the growth rate. In this study, the growth-government size
relationship will be tested additionally with a quantile regression model. The extra insight
from quantile regression model is that it can reveal whether the effect of government size
depends on the percentile (quantile) of the distributions of the variables. It could be that the
growth impact of government size will be different depending on whether an economy is
growing slower or faster than average. Quantile regression will allow us to obtain a more
comprehensive picture. Chen et al. (2011) claim that the method had not been applied to the
growth-government size relationship prior to their study. This current study may be the
second one to apply the method to the issue using a different dataset.
2.2 Decomposition of Nonstationary Economic Time Series
Any nonstationary (trending) time series can be decomposed into permanent (or ‘trend’) and
transitory (or ‘cycle’) components (Garratt et al., 2006). The decomposition principle entails
applying some pre-filtering procedure, usually univariate in nature, to extract a trend (or
permanent component). The trend could be deterministic or stochastic. The deviations about
the trend constitute the detrended series or transitory/cycle component that is assumed to be
stationary. The trend component captures shocks that have a permanent effect on the level of
the variable, and the cycle component captures shocks that only have a temporary effect on
the level of the variable. There is no unique way to decompose a time series into trend and
cyclical parts. The main competing decomposition techniques are Beveridge and Nelson
(1981) model, the unobserved components models of Harvey (1985) and Clark (1987), the
business cycle filters of Hodrick and Prescott (1997) and Baxter and King (1999), and some
others such as wavelet transformation (Padda, 2011) and phase average trend (Boschan and
Ebanks, 1978; Zarnowitz and Ozyildirim, 2006). The decomposition techniques are
distinguished by how they define the trend and cycle components, and there are numerous
combinations of these. For instance, the Beveridge-Nelson and Harvey-Clark methods define
the trend as a random walk with drift but they differ in how they define serial correlation in
the model. Since the detrending filters used by the various techniques extract different kinds
of information from the data under consideration, the decomposition results vary and each
method suffers from significant deficiencies (Padda, 2011). The Hodrick-Prescott (HP)
method, however, has survived a lot of tests and is now perhaps the most popular.4 Hence the
HP decomposition method is adopted in this study. 4 Banerji and Dua (2011) report that the OECD has switched from phase average trend (PAT) method to the HP
as their decomposition method in growth cycle analysis. For its growth cycle, Statistics New Zealand employs
HP, Baxter-King and the Henderson filters (see Statistics New Zealand, 2007).
6
3. METHODOLOGY AND DATA
3.1 The Analytical Methods
3.1.1 The Growth-Government Size Model
To check the growth-government size relationship in the data set, this study adopted the
growth equation expounded by Dar and AmirKhalkhali (2002) and which was utilised by
Chen et al. (2011). Extending the arguments in the Solow [exogenous] growth accounting
method which postulates that the rate of economic growth is a function of the growth in total
factor productivity (TFP) and the weighted growth rates of capital and labour, Dar and
AmirKhalkhali include government size and the growth rate of exports to endogenise the
growth model. That is,
gY = f(gK, gL, gEX, GS) (1)
where gY is growth rate of GDP, gK is growth rate of capital, gL is growth rate of the labour
force and GS is government size. Dar and AmirKhalkhali argue that ‘export growth (through
its favourable impact on the efficiency of resource use, innovative activity and the rate of
technical progress, and the realisation of economies of scale) raises TFP growth and, by
implication, economic growth.’ Government size, like export expansion, impacts on
economic growth through its impact on either total or individual factor productivity. In the
literature, some studies prefer to use growth in the per capita GDP to represent economic
growth in which case the per-capita equivalent of the growth model may be written as:
gy = f(gk, gEX, GS) (2)
where lower case y and k represent per capita GDP and capital per capita, respectively.
For the empirical estimation, panel regression techniques rather than pooled OLS were
employed because the latter is biased (from omitting a time-constant variable for the
countries) and is also inconsistent if the time-constant variable and the explanatory variables
are correlated (Wooldridge, 2013, pp. 484-496). The specific counterparts of equations (1)
and (2) to be estimated in this study can be written, respectively, as equations (3) and (4):
GRYit = β0 + β1GRKit + β2GRLit + β3GREXit + β4GOVSIZEit + ui + eit (3)
GRYPCit = γ0 + γ1GRKPCit + γ2GREXit + γ3GOVSIZEit + ui + eit (4)
for i = 1, …, 24; t = 1, …, 39, and where GRY is growth rate of GDP, GRK is the growth rate
of gross capital stock, GRL is the growth rate of the labour force, GREX is rate of export
expansion, GOVSIZE is government final consumption as percentage of GDP, GRYPC is the
growth rate of per capita GDP, GRKPC is the growth rate of capital per capita, ui is the
unobserved heterogeneity or the fixed effects for country i that is assumed to be correlated to
all explanatory variables at all times, and eit represents the idiosyncratic error or unobserved
factors that change over time and affect growth. To check for the existence of the BARS
curve (or nonlinear relationship) and potentially estimate an optimal value of government
size, the quadratic term for government size can be added to each of equations (3) and (4).
The main reason for implementing panel estimation is to allow for the unobserved effect ui to
7
be correlated with the explanatory variables. That is, we want to allow the unmeasured
country-specific time-invariant factors that affect growth to also be correlated with
government size. The intercept reported in the panel model is the average across the countries
of the estimated country-specific intercepts. However, individual country intercepts can be
estimated with an appropriate formula and they show whether the unobserved fixed effects
that contribute to growth are above or below the average in the sample (Wooldridge, 2013, p.
489). The Hausman test would be used to choose between random effects and fixed effects.
Quantile regressions of equations (3) and (4) were also estimated to check whether the
regression parameters changed with quantiles of the distribution of the dependent variable.
The quantile regression methodology is described by Koenker (2005) and Koenker and
Bassett (1978).
3.1.2 The Hodrick-Prescott Decomposition Model
As mentioned in Section 2.2, the competing decomposition techniques are distinguished by
how they define the trend and cycle components. A brief description of the HP technique is
offered here. Let a time series yt be viewed as the sum of a trend component τt and a cyclical
component ct. That is,
yt = τt + ct for t =1,…., T. (5)
The HP method of extracting τt requires the minimization of the cost
∑ ( ) ∑ ( ) ( )
(6)
where the first term penalises the variance in the cyclical component while λ > 0 penalizes
variability in the trend component and therefore describes how much the lack of smoothness
in the trend contributes to the overall cost. The larger λ is, the smoother is the trend
component. For annual data, Hodrick and Prescott recommend a value of λ = 100.
3.2 The Data
All the data were sourced from the websites of the World Bank. Table 1 reports the averages
of the variables of interest for the sampled countries. Since the main focus of the study is on
government size, that variable is the one that will be described in some detail in this section.
The raw panel data on government size are presented in Appendix 1. The values of
GOVSIZE ranged from a minimum of 6.1% (posted by Turkey in 1988) to a maximum of
29.8% (posted by Denmark in 2009), averaging 18.95%. The country whose period average
comes closest to it is Austria, with 18.7%. The country with the highest sample average is
Sweden (26.9%) followed by Denmark (25.9%) and Netherlands (24.0%); and the country
with the lowest sample average is Turkey with 10.1%. New Zealand and Australia have
period averages of 18.3% and 17.7%, respectively. The year 2009 was when the sample
annual average peaked (at 21.6%) and the minimum yearly average occurred in 1973 (at
15.6%). Consistent with the upward trend of government size in the sample period, the range
of values at the start of the study period was from 8.6% to 23% and at the end of the study
period it was from 11.1% to 28.6%. In the sample, labour force, capital stock and capital per
worker grew at the average rates of 0.7%, 2.9% and 2.2% per annum, respectively. For the
8
purposes of the quantile regressions, the summary statistics and the bar charts of the
distributions of GRYPC (growth rate of per capita GDP) and GRY (growth rate of aggregate
GDP) are presented in Figure 1. Both proxies of economic growth are slightly skewed to the
left. GRY ranged from -8.5% to 11.2% and averaged 2.6%; GRYPC ranged from -9.0% to
11.2% and averaged 1.9%.
To check the extent of co-movements of the government size variable between countries, the
cross-country correlations were estimated. Appendix 2 reports the pair-wise correlations of
government size between countries. Of the 276 calculations, about 81% are positive and 19%
negative, suggesting a high incidence of co-movement among the series. The highest
estimated correlation is 0.97 between Iceland and Portugal. Other very highly correlated
countries are France and Spain (0.93), Finland and Spain (0.92) and Iceland and Japan (0.90).
New Zealand is most highly correlated with Netherlands (0.75), second with Denmark (0.71)
and least with Iceland (0.16). The correlation with Australia is moderate at 0.51.
Table 1
Basic Average Statistics for the 24 OECD Countries (%)
Country GRY GRYPC GRK GRKPC GRL GREX GOVSIZE
AUS 3.15 1.79 4.71 3.23 1.48 4.90 17.68
AUT 2.38 2.09 2.16 1.79 0.36 5.33 18.72
BEL 2.14 1.81 2.54 2.07 0.47 4.22 22.01
CAN 2.83 1.67 4.28 3.07 1.21 4.05 21.01
DNK 1.80 1.51 2.34 2.16 0.19 4.86 25.85
FIN 2.59 2.20 2.61 2.18 0.43 5.38 21.13
FRA 2.14 1.59 1.99 1.39 0.60 4.75 22.54
GER 1.99 1.89 1.48 1.27 0.21 5.77 19.78
GRC 1.89 1.27 0.40 -0.13 0.52 5.56 16.22
ISL 3.17 2.09 4.06 3.07 0.99 4.47 21.08
IRL 4.37 3.31 2.96 1.90 1.06 9.03 18.55
ITA 1.99 1.70 1.93 1.60 0.33 4.36 18.61
JPN 2.48 2.03 1.35 0.98 0.36 5.96 15.54
LUX 3.78 2.75 4.89 3.72 1.18 5.79 15.81
NLD 2.39 1.82 2.05 1.35 0.71 5.03 23.96
NZL 2.32 1.35 3.37 2.21 1.16 3.96 18.30
NOR 2.95 2.36 2.88 2.21 0.66 3.83 20.40
PRT 2.59 2.08 2.94 2.55 0.39 4.99 16.28
ESP 2.63 1.88 2.48 1.64 0.84 6.09 16.34
SWE 2.19 1.81 2.58 2.16 0.42 5.03 26.90
CHE 2.18 1.58 1.17 0.58 0.59 4.21 10.85
TUR 4.28 2.52 8.16 6.40 1.76 9.25 10.09
GBR 2.19 1.90 2.73 2.47 0.26 4.28 20.61
USA 2.77 1.76 3.16 2.29 0.87 6.04 16.45
All 2.63 1.95 2.88 2.17 0.71 5.30 18.95
Sources: World Bank and OECD databases.
9
Figure 1
Frequency Distributions and Summary Statistics of the Growth Variables
4. ANALYTICAL RESULTS
4.1 Impact of Government Size and the BARS Curve
Before decomposing the government size time series, the effect of government size on
economic growth in the dataset was checked. In connection with this, two models were
estimated. In the first one, the dependent variable is GRYPC – the growth rate of per capita
GDP (equation (6)); and in the second one, the dependent variable is GRY – the growth rate
of GDP (equation (5)). All variables were expressed in percentages. Each model was then
augmented with the quadratic term of the government size variable to check for the existence
of the BARS curve. In the growth panel regressions, the Hausman test rejected the random
effects in preference for the fixed effects. The fixed effects results are therefore reported in
Table 2.
The results of the linear models in Table 2 (Models 1.1 and 2.1) indicate that government size
is significantly negatively related to the growth rates of both GDP and per capita GDP. This
corroborates the results of the majority of empirical studies as mentioned in Section 2. An
increase of one percentage point in government size is expected to decrease the growth rates
of GDP and per capita GDP by about 0.13 and 0.11 percentage points, respectively.
Illarionov and Pivarova (2002) studying the OECD countries over the 1960-2000 period
estimated that a rise of one percentage point in the share of public expenditure on GDP came
with a reduction of 0.1 percentage point in the average growth rate of economic activity.
Other studies on the OECD/EU have reported similar findings.5 The capital, labour and
export variables take the expected positive signs in all the models and they are all very
significant.
5 For instance, Gwartney et al. (1998), Folster and Henrekson (2001), Pevcin (2004) and Afonso and Furceri
(2008).
10
Table 2
Panel (Fixed Effects) Regression Results
Variable
Dependent Variable is GRYPC Dependent Variable is GRY
Model 1.1
(Linear)
Model 1.2
(Quadratic)
Model 2.1
(Linear)
Model 2.2
(Quadratic)
Constant 3.1591***
(4.96)
1.5565
(1.22)
3.9236***
(6.27)
3.5324***
(2.86)
GRKPC 0.1492***
(26.94)
0.1488***
(26.86)
GRK 0.1514***
(28.55)
0.1513***
(28.49)
GRL 0.1024***
(3.38)
0.1027***
(3.39)
GREX 0.1061***
(11.61)
0.1076***
(11.70)
0.1081***
(12.36)
0.1085***
(12.31)
GOVSIZE -0.1107***
(-3.33)
0.0738
(0.56)
-0.1252***
(-3.87)
-0.0803
(-0.64)
GOVSIZE_SQR -0.0050
(-1.44)
-0.0012
(-0.37)
Statistics
R2 0.7050 0.7057 0.7386 0.7387
Adj R2 0.6833 0.6837 0.7191 0.7188
No. of obsvns 936 936 936 936 Notes: Figures in parentheses are t-statistics. Triple, double and single asterisks (i.e., ***, ** and *)
indicate statistical significance at the 1%, 5% and 10% levels, respectively.
A BARS curve exists if the government size variable takes a positive sign and its quadratic
term takes a negative sign. The introduction of the quadratic term of government size points
to the existence of an inverted-U relationship (and therefore the existence of the BARS curve)
in the whole sample when growth rate of per capita GDP is used to represent economic
growth (Model 1.2) but not when economic growth is represented with GDP growth rate. It is
not clear why this is the case. It would seem from this result that for analytical purposes
economic growth is better captured with per capita GDP growth rate than with GDP growth
rate.6 Although the quadratic relationship is not statistically significant, it is estimated that the
optimal government size for the whole sample is 7.38%. Previous studies have confirmed the
existence of the BARS curve for various countries using growth rates of either GDP or per
capita GDP: Vedder and Galloway (1998) for the US, Canada, Denmark, Italy, Sweden and
Britain; Handoussa and Reiffers (2003) for Tunisia; Pevcin (2004) for 12 European countries;
Radwan and Reiffers (2004); Chen and Lee (2005) for Taiwan; Chiou-Wei et al. (2010) for
South Korea, Taiwan and Thailand; De Witte and Moesen (2010) for 23 OECD countries
using the nonparametric data envelopment analysis (DEA); and Herath (2012) for Sri Lanka.
6 Herath (2012) explains that real GDP is an aggregate figure that does not account for differing sizes of nations
and the better measurement of actual economic growth is per capita income because it reflects the average
standard of living of individual members of the population.
11
Compared with the group period average of 18.95%, the estimated optimal government size
of 7.38% shows that the typical OECD country exceeded the threshold level by more than
two-and-a-half times. Because some past studies used government total expenditure rather
than final consumption expenditure, the estimate of the optimal government size can only
truly be compared with those studies that utilised the share of government final consumption
in GDP to represent government size.7 In that regard, Chobanov and Mladenova (2009)
estimated an optimum government size of 10.4% based on 1961-2005 data for a panel of 81
countries. Owing to model and data limitations, the authors thought their estimate was
probably biased upwards. Davies (2008) studied the panel dataset on 154 countries over the
period 1975 through 2002 and estimated that the growth-maximising government size was
about 8.5%. Chiou-Wei et al. (2010), studying South Korea, Malaysia, Singapore, Taiwan
and Thailand and using the same indicators for government size and economic growth as this
current study, estimated that the threshold of government size for those countries (except
Malaysia) was about 11% although the individual country threshold levels were different.
The authors did not find nonlinearity in the Malaysian data. It must be remembered that the
data used by Chiou-Wei et al. (2010) covered the 1961-2004 period and the authors
implemented the STAR model.
Following the confirmation of the existence of the BARS curve in the consolidated data when
GDP per capita growth rate is the dependent variable, efforts were made to find out if the
curve existed for individual countries. The OLS results for the individual countries and the
estimated turning points of government size are reported in Table 3. It will be seen in Table 3
that the BARS curve (the inverse-U relationship) is confirmed for only 13 out of the 24
countries. For the rest of the countries, the relationship was estimated to be U-shaped, which
contradicts the BARS curve. However, there is a precedence of this in the literature, as it will
be disclosed shortly. Among the 13 countries for which the BARS curve was confirmed, the
estimated optimal government size ranged from 11.04% (for TUR) to 23.10% (for DNK).
Eleven of these countries (AUS, CAN, DNK, FIN, FRA, ISL, ITA, LUX, NZL, SWE and
GBR) had period averages larger than their estimated optimal government sizes. The inverse-
U relationship was strong in the cases of CAN, FIN, NZL and GBR but weak in the rest. For
these four countries (Canada, Finland, New Zealand and Great Britain), therefore,
downsizing could be growth-enhancing. The two countries that had period averages smaller
than the estimated optimal government sizes – Portugal (PRT) and Turkey (TUR) – could
presumably benefit from increases in government size. It is worthy of note that for four of the
five Asian countries in their sample, Chiou-Wei et al. (2010) estimated the turning point in
government size occurred at about 11% for Korea, Singapore and Thailand and about 16%
for Taiwan.8 For Singapore, however, the relationship was estimated to be U-shaped, in
contradiction to the BARS curve. The authors explained that, relative to the other countries,
Singapore’s government was strong and efficient enough to positively influence economic
growth even when it got bigger than 11%. In the current study, however, it is surmised that a
U-shaped relationship indicates that spending cuts would crimp growth rather than boost it.
7 Studies that utilised total government expenditure as ratio of GDP have typically reported optimal government
size ranging from 20% to 30% (Chobanov and Mladenova, 2009). 8 In an earlier study, Chen and Lee (2005) had estimated a threshold level of about 15% for Taiwan.
12
Table 3
Results of Regressions to Estimate the Optimal Government Size for the Sample Countries
Country Independent Variable
R2 Shape
Turning
Point
Period
Avge Constant GRKPC GREX GOVSIZE GOVSIZE2
AUS -23.367
(-1.07)
0.1976
(9.41)
0.1051
(3.11)
3.0134
(1.10)
-0.0935
(-1.10) 0.7504 ∩ 16.12% 17.7%
AUT 18.3662
(0.97)
0.1519
(4.83)
0.1058
(2.11)
-1.5027
(-0.70)
0.0314
(0.52) 0.6723 U 23.96% 18.7%
BEL 10.7938
(0.52)
0.1031
(3.01)
0.2048
(4.70)
-0.6825
(-0.35)
0.0102
(0.23) 0.8230 U 33.49% 22.0%
CAN -65.48
(-2.17)
0.1134
(4.96)
0.2005
(7.53)
6.7738
(2.35)
-0.1721
(-2.51) 0.8522 ∩ 19.68% 21.0%
DNK -19.772
(-1.46)
0.1839
(13.07)
0.0829
(3.08)
1.8062
(1.69)
-0.0391
(-1.85) 0.8672 ∩ 23.10% 25.9%
FIN -33.977
(-1.73)
0.1827
(8.19)
0.0965
(2.52)
3.8521
(2.06)
-0.1020
(-2.31) 0.8539 ∩ 18.88% 21.1%
FRA -0.2196
(-0.02)
0.1464
(8.78)
0.1134
(6.23)
0.5573
(0.62)
-0.0225
(-1.07) 0.9221 ∩ 12.39% 22.5%
GER 82.5186
(1.73)
0.1984
(11.61)
0.1743
(8.19)
-8.5865
(-1.77)
0.2243
(1.83) 0.8321 U 19.14% 19.8%
GRC 26.2715
(2.36)
0.2154
(7.30)
0.0795
(2.45)
-3.2662
(-2.42)
0.1032
(2.48) 0.6338 U 15.83% 16.2%
ISL -12.587
(-0.80)
0.1318
(8.06)
0.3633
(7.61)
1.4570
(0.98)
-0.0398
(-1.15) 0.7632 ∩ 18.31% 21.1%
IRL 24.9305
(1.20)
0.1096
(2.78)
0.3389
(5.77)
-2.4103
(-1.05)
0.0569
(0.91) 0.7243 U 21.17% 18.5%
ITA -8.6719
(-0.63)
0.2033
(16.00)
0.0961
(5.17)
1.6279
(1.06)
-0.0593
(-1.40) 0.8967 ∩ 13.72% 18.6%
JPN 15.2545
(1.20)
0.3491
(5.86)
0.0324
(0.95)
-1.5278
(-0.98)
0.0405
(0.86) 0.8533 U 18.87% 15.5%
LUX -8.7122
(-0.38)
0.0889
(2.42)
0.3483
(5.92)
1.5566
(0.46)
-0.0617
(-0.51) 0.6854 ∩ 12.62% 15.8%
NLD 20.2550
(1.40)
0.1390
(5.79)
0.1658
(6.68)
-1.2086
(-1.07)
0.0165
(0.76) 0.8537 U 36.71% 24.0%
NZL -51.407
(-1.63)
0.1036
(3.92)
0.1765
(2.21)
6.1754
(1.77)
-0.1819
(-1.90) 0.5986 ∩ 16.98% 18.3%
NOR 15.2046
(0.33)
0.1097
(5.61)
0.3257
(5.85)
-1.1891
(-0.26)
0.0237
(0.22) 0.6619 U 25.05% 20.4%
PRT 14.7010
(-0.19)
0.0228
(7.15)
0.0276
(3.11)
1.7922
(0.36)
0.0527
(-0.45) 0.7306 ∩ 17.01% 16.3%
ESP 4.8187
(1.54)
0.2949
(14.24)
0.1044
(3.60)
-0.3480
(-0.82)
0.0060
(0.43) 0.8587 U 29.23% 16.3%
SWE -10.631
(-0.35)
0.1358
(5.61)
0.1566
(3.67)
1.1206
(0.48)
-0.0259
(-0.58) 0.7732 ∩ 21.65% 26.9%
CHE 80.5783
(3.57)
-0.0009
(-0.01)
0.1605
(2.26)
-13.4694
(-3.03)
0.5610
(2.60) 0.5453 U 12.00% 10.9%
TUR -3.5814
(-0.71)
0.1988
(13.61)
0.0251
(1.39)
0.8744
(0.83)
-0.0396
(-0.75) 0.8307 ∩ 11.04% 10.1%
GBR -63.272
(-2.31)
0.1985
(8.73)
0.0274
(0.70)
6.8236
(2.55)
-0.1783
(-2.75) 0.8422 ∩ 19.14% 20.6%
USA 49.1177
(1.47)
0.2205
(11.79)
0.0658
(2.44)
-5.9348
(-1.40)
0.1817
(1.35) 0.8660 U 16.33% 16.5%
Notes: Figures in parentheses are t-statistics generated from HAC robust standard errors.
13
Of the 11 countries in the current study for which the government size-growth relationship
was U-shaped, 8 of them (AUT, BEL, IRL, JPN, NLD, NOR, ESP and CHE) had period
averages smaller than their respective estimated turning points of government size. The
relationship was strong for CHE but weak for the others. They may be said to be on the
declining part of the parabola where increases in government size would mitigate the
negative impact until government size reaches the estimated turning point. After that, further
increases in government size are expected to have positive impact on growth. The remaining
three countries, having period averages larger than their respective turning point values
(GER, GRC and USA), may be said to be on the rising segment of the parabola. This result
suggests those countries may increase government size without adversely affecting economic
growth. The relationship was strong for GER and GRC but weak for USA. In summary, the
findings about the growth-government size relationship for individual countries confirm a
robust BARS curve for CAN, DNK, FIN, NZL and GBR, a significant U-shaped relationship
for GER, GRC and CHE and insignificant relationship in the other countries.9
4.2 Quantile Regression Results
The results of the quantile regressions for selected quantiles are reported in Tables 4 and 5,
and the corresponding 95% confidence intervals of the GOVSIZE coefficients are plotted in
Figure 2. As the quantile regression results in Table 4 show, the impact of government size
on the per capita GDP growth rate is positive and insignificant below the 40th
quantile and
becomes negative from the 40th
through the 95th
quantile. For GDP growth rate, the impact is
also positive and insignificant below the 20th
quantile after which it becomes negative and
increasingly significant through the 95th
quantile (see Table 5).10
These results corroborate
the finding by Chen et al., (2011), who first reported such differential impact of government
size at different quantiles from a study of an unbalanced panel dataset for 24 OECD countries
for the 1971-2001 period. That study used GDP growth rate as the indicator for economic
growth and the ratio of government total expenditure to GDP as the proxy for government
size. The current study’s results add to the literature by showing that qualitatively similar
differential impacts are obtained for the alternative proxy for government size – government
final consumption as a ratio of GDP – when either economic growth indicator is used. The
implication from the analytical result from the current study is that when the per capita GDP
growth rate is less than 1.76% per annum or the aggregate GDP growth rate is less than 2.6%
per annum, [well targeted] stimulus can be growth-enhancing.
9 Re-estimation of the country models augmented with a trend variable did not appreciably change the overall
results. It was noticed, however, that the trend variable took a negative and significant coefficient (at the 5%
level or lower) in the cases of BEL, CAN, DNK, GER, ITA, LUX, NOR and USA, and at the 10% level for
IRL. In the cases of ESP and TUR the trend variable took a positive and significant coefficient at the 1% and
10% levels, respectively. The presence of the trend variable also forced the government size coefficients for
CAN to be insignificant and those for BEL and ESP to be monotonically decreasing and insignificant. 10
For the model with per capita GDP growth rate (GRYPC) as dependent variable, the sign of the coefficient is
positive up to the 14th
quantile and turns negative afterwards; the corresponding quantile for the model with
GDP growth rate (GRY) is the 36th
. Given the probability distributions of these variables in Figure 1, and
assuming normal distribution, the cut-off growth rates are 1.76% and 2.60%, respectively, for per capita GDP
growth rate and GDP growth rate.
14
Table 4
Quantile Regression Results – Dependent Variable is GRYPC
Variable Quantile, τ
0.10 0.20 0.30 0.40 0.50 0.75 0.95
Constant -1.0509***
(-2.67)
-0.4238
(-1.11)
-0.0759
(-0.31)
0.7419***
(2.80)
1.2977***
(5.80)
3.1892***
(6.88)
7.1354***
(11.35)
GRKPC 0.1721***
(18.01)
0.1703***
(20.93)
0.1684***
(27.08)
0.1651***
(21.99)
0.1706***
(20.79)
0.1616***
(16.20)
0.1122***
(8.87)
GREX 0.1237***
(7.17)
0.1302***
(7.51)
0.1365***
(9.54)
0.1429***
(12.70)
0.1403***
(16.61)
0.1506***
(9.22)
0.1680***
(6.43)
GOVSIZE 0.0123
(0.64)
0.0113
(0.62)
0.0119
(0.97)
-0.0146
(-1.17)
-0.0280***
(-2.58)
-0.0760***
(-3.40)
-0.1969***
(-7.58)
Statistics
Pseudo-R² 0.4533 0.4359 0.4205 0.4004 0.3832 0.3590 0.3805
Adj R² 0.4516 0.4340 0.4186 0.3985 0.3812 0.3570 0.3786
No. of Obs 936 936 936 936 936 936 936
Notes: Figures in parentheses are t-statistics which were obtained from the bootstrapped standard errors. Triple,
double and single asterisks (i.e., ***, ** and *) indicate statistical significance at the 1%, 5% and 10% levels,
respectively.
Table 5
Quantile Regression Results – Dependent Variable is GRY
Variable Quantile, τ
0.10 0.20 0.30 0.40 0.50 0.75 0.95
Constant -0.2750
(-0.97)
0.4816
(1.55)
1.0991***
(4.67)
1.5056
(6.08)
2.2413***
(7.12)
4.2814***
(9.96)
8.1099***
(13.91)
GRK 0.1685***
(15.95)
0.1641***
(21.09)
0.1609***
(21.83)
0.1605***
(21.41)
0.1699***
(18.18)
0.1594***
(19.93)
0.1183***
(6.73)
GRL 0.1327**
(2.24)
0.1846***
(3.66)
0.1949***
(4.39)
0.1938***
(5.12)
0.1289***
(3.16)
0.1279**
(2.10)
0.1387*
(1.83)
GREX 0.1351***
(6.66)
0.1262***
(8.27)
0.1254***
(9.64)
0.1275***
(12.13)
0.1225***
(12.01)
0.1398***
(8.93)
0.1764***
(5.96)
GOVSIZE 0.0002
(0.02)
-0.0114
(-0.83)
-0.0234**
(-2.36)
-0.0302***
(-2.71)
-0.0505***
(-3.60)
-0.1102***
(-5.70)
-0.2296***
(-9.14)
Statistics
Pseudo-R² 0.5093 0.4895 0.4707 0.4531 0.4352 0.3998 0.4169
Adj R² 0.5072 0.4873 0.4684 0.4508 0.4327 0.3972 0.4144
No. of Obs 936 936 936 936 936 936 936
Notes: Figures in parentheses are t-statistics which were obtained from the bootstrapped standard errors. Triple,
double and single asterisks (i.e., ***, ** and *) indicate statistical significance at the 1%, 5% and 10% levels,
respectively.
15
Figure 2
Plots of the Point Estimates and 95% Confidence Intervals of the Coefficient for the
Government Size Variable for all Quantiles for the Two Growth Models
4.3 Government Size Time Series Decomposition Results
Prior to the decomposition of the government size time series, their degrees of integration
were checked to ensure they were nonstationary and therefore decomposable. To check the
stationarity of the government size time series for each country, the Augmented Dickey
Fuller (ADF) and the Phillips Perron (PP) tests for unit root were implemented. The results of
the two tests were comparable and for brevity the ADF results are reported in Table 6. The
series for all the countries (and for the OECD group and the World11
) are integrated of order
one (i.e., I(1) or nonstationary) except for Austria, Germany and Iceland. For the latter two
countries the series are trend-stationary; only the series for Austria is stationary. Hence for
practical purposes it was assumed that all the series are nonstationary and therefore amenable
to decomposition.
Figure 3 shows plots of the observed government size time series and the HP trend and cycle
components for the sampled countries. In each illustration, the observed series is plotted with
diamond markers, the HP trend is plotted as a smooth solid line and the HP cycle is plotted as
a dashed line. It will be seen that the trend line tracks the actual series quite closely in all
countries. There are noticeable cycle components where the actual series diverge from the
trend. A visual examination suggests that the duration and dates of ‘government business
cycles’ are not the same for all the sampled countries. Since the objective of the study is not
to describe and date cycles, attention will be turned to the relative importance of the trend and
cycle components.
11
The results for the OECD group and for the world are included for comparison.
16
Table 6
Results of the ADF Unit Root Tests on Government Size
Country Constant Trend p-value Conclusion AUS Yes No 0.0135 I(1) AUT Yes No 0.0040 I(0) BEL Yes No 0.0995 I(1) CAN Yes No 0.1459 I(1) DNK Yes No 0.0698 I(1) FIN Yes Yes 0.0791 I(1) FRA Yes Yes 0.0265 I(1) GER Yes Yes 0.0043 I(0) GRC Yes Yes 0.0308 I(1) ISL Yes Yes 0.0031 I(0) IRL Yes No 0.6001 I(1) ITA Yes No 0.6703 I(1) JPN Yes No 0.9839 I(1) LUX Yes No 0.0095 I(1) NLD Yes No 0.6488 I(1) NZL Yes No 0.2566 I(1) NOR Yes No 0.0721 I(1) PRT Yes No 0.4587 I(1) ESP Yes No 0.2634 I(1) SWE Yes No 0.0138 I(1) CHE Yes No 0.0652 I(1) TUR Yes Yes 0.3015 I(1) GBR Yes No 0.1021 I(1) USA Yes No 0.2633 I(1) OECD Yes Yes 0.0232 I(1) World Yes No 0.3408 I(1)
The relative importance of the permanent and transitory components are determined by which
one accounts for more of the observed variance in the series (Murray and Papanyan, 2004).
The estimated variances of the HP trend and cycle components for the sampled countries are
reported in Table 7. For all countries except Australia, the trend variance is several times
bigger than the cycle variance and therefore the trend accounts for more of the variance in the
observed series. This suggests that permanent shocks are relatively more important than
transitory shocks. That is, most of the interesting dynamics of the government size series are
captured by the trend component and the cycle is largely noise. The permanency of the
changes is strongest in Portugal, followed by Spain, Japan and Iceland; it is weakest in
Luxembourg. It is only in Australia that the transitory changes trump the permanent changes.
17
Figure 3
Plots of Observed Government Size Series and Estimated HP Trend and Cycle Components
18
19
20
Table 7
Estimated Variances of the HP Trend and Cycle Components
Country Trend Variance Cycle Variance Country Trend Variance Cycle Variance
AUS 0.2563 0.2711 IRL 2.9297 0.7769
AUT 0.7254 0.1970 ISL 9.5515 0.3963
BEL 1.1087 0.3895 ITA 1.7351 0.2599
CAN 1.0000 0.6912 JPN 4.7805 0.1816
CHE 0.6359 0.1042 LUX 0.5614 0.5320
DNK 1.2487 0.7110 NLD 1.9550 0.4049
ESP 6.9425 0.2557 NOR 0.9900 0.8050
FIN 4.5203 0.8626 NZL 0.6832 0.4013
FRA 2.2196 0.2595 PRT 10.9204 0.3145
GBR 1.0420 0.5831 SWE 0.8759 0.7937
GER 0.5809 0.1967 TUR 4.1727 0.8018
GRC 3.0230 0.6844 USA 0.7762 0.1930
To explain why the changes in government size are predominantly of the permanent rather
than transitory nature, one needs to take a look at the extent of regulations within the
economy and the role of government in providing goods and services. Everywhere, the
increasing concerns about safety, financial crises and the environment have induced
corresponding expansions in the regulatory framework. The two important areas in which
government is a dominant player in service delivery are education and health. OECD reports
that, on the average, governments account for 70% and 85% of the final consumption
expenditures on health and education, respectively (OECD, 2011). With positive rates of
population growth and longer life expectancy, governments are faced with non-decreasing
demand for education and health services. For instance, schooling is compulsory until at least
the age of 15 and the majority of primary and secondary students are enrolled in government
run/financed institutions. The natural rate of population increase, no matter how low, implies
greater outlay to provide basic education to the young. Longer life expectancy and the
looming retirement of the ‘baby boomers’ means aged care will feature more prominently in
health services. This will be particularly acute in countries with universal public health
insurance.
The finding or conclusion from the HP decomposition that the changes in government size
are predominantly of the permanent rather than transitory nature warrants disambiguation
from Wagner’s law of increasing state activity and the Peacock-Wiseman hypothesis (in
public finance) both of which emphasise that public expenditure has a tendency to increase
over time and with economic development (Singh, 2008). Wagner’s law states that as nations
industrialise, the share of the public sector in the national economy grows continually for
reasons such as the state social functions that expand over time, administrative and protective
functions and welfare functions (Wagner, 1893, 1911). The Peacock-Wiseman hypothesis,
which is an elaboration of Wagner’s Law, states that over the years economic development
and income growth and the concomitant enlarged tax base results in an increase in
government revenue. That, in turn, leads to a boost in public expenditure because taxpayers
demand the provision of various services which the government cannot ignore (Peacock and
21
Wiseman, 1961). Wagner’s Law is validated when the long-run elasticity of government
expenditure with respect to GDP is larger than 1, and it is violated when the elasticity is less
than 1. In a review of the empirical assessments of Wagner’s Law, Durevall and Henrekson
(2011) report that around 65% of the studies find direct or indirect evidence in favour of the
concept while 35% provide no support. In a recent study on 23 OECD economies over the
period 1970-2006, Lamartina and Zaghini (2012) confirmed the existence of Wagner’s Law
but noted that the elasticity has declined over the years and also it is smaller for the relatively
richer countries. The latter observation is similar to the finding by Kuckuck (2012) who
tested the validity of Wagner’s Law for the UK, Denmark, Sweden, Finland and Italy over
the period 1850-2010 (more than 160 years). She found that the relationship between public
spending and economic growth has weakened with economic advancement and may have
reached its limit in recent decades.
The simultaneous consideration of the BARS curve and Wagner’s Law leads to what
Balatsky (2012) has referred to as ‘the paradox of wealth’. If, according to Wagner’s Law,
government size will continually increase with economic growth/development but
government size beyond a certain level has a dampening effect on the rate of growth, then it
is conceivable that if government size gets big enough it could stymie economic growth and
cause stagnation of wealth creation. Given the predominantly democratic institutions in
OECD countries, for instance, that level of displacement of the private sector by the public
sector is not likely to eventuate. In the event, Wagner’s Law acts as a serious limit to an
economy’s long-term growth and to eliminate the paradox of wealth it is necessary to
neutralise or violate that law. Findings of weakened relationship between government
spending and economic growth in the wealthiest countries and over time in growing
economies (e.g., Lamartina and Zaghini, 2012; Kuckuck, 2012) attest to the decline of
Wagner’s Law as part of economic evolution with its attendant growth. The consequence is
that the public sector’s share has ceased to increase monotonically and has changed over to a
fluctuating mode dynamically competing with the private sector. The finding that the shocks
to government size leave mainly permanent effects, rather than transitory effects, suggests
that deliberate policy choices have to be made to reduce government size. And here, the
education and health sectors offer perhaps the best leverage. In health, reforms that can
decrease the role of government include co-payments by households for some services like
doctor visits, privatisation of state-owned hospitals and the liberalisation of the health
insurance market. In the expenditure on education, OECD (2011) reports that most of the
differences between countries lie in the extent to which the governments finance pre-primary
and tertiary education. Korea, for instance, has a relatively higher enrolment rate in private
educational institutions at the pre-primary and university levels as well as a higher use of
private tutoring. Thus, Korea’s public expenditure on education as a percentage of GDP is
one of the lowest in the OECD.
5. SUMMARY AND CONCLUSION
The concurrence of large government size, slow growth and persistent unemployment in the
wake of the 2008/09 global financial crisis and the widespread expectation for governments
to downsize has renewed interest in the relationship between government size and economic
growth. The study analysed four aspects of the fluctuations in government size in 24
countries that have been members of the OECD since 1973; the most up to date data were
used. Firstly, the nature of the relationship between government size and economic growth
22
was investigated. A check with panel fixed effects model found that economic growth is
significantly negatively related to government size. Secondly, the introduction of the
quadratic term of the government size variable revealed that the relationship has an inverse-U
shape for the whole sample which provided a basis for the estimation of a growth-maximising
size of government. Checks of the individual sampled countries’ data yielded almost an equal
mixture of countries with U-shaped and inverse-U shaped relationships. This suggests that
increases in government size are not uniformly detrimental to economic growth for all
countries. Thirdly, a further check with quantile regression models for the whole sample
revealed that the impact of government size on economic growth is positive and insignificant
at low rates of economic growth. This impact decreases as economic growth increases and
eventually turns negative and significant at relatively high rates of economic growth. This
finding corroborates the finding by the first study to apply quantile regression to the
government size-economic growth relationship (Chen et al., 2011) that analysed an
unbalanced panel dataset for 24 OECD countries for the 1971-2001 period.
Fourthly, the study decomposed the government size time series of the sampled countries into
the permanent/trend and transitory/cycle components. This was done in an effort to
characterise whether the shocks to government size are predominantly of the permanent type
or of the transitory type. The trend component captures shocks that have a permanent effect
on the level of the variable, and the cycle component captures shocks that only have a
temporary effect on the level of the variable. Permanent shocks might need deliberate counter
measures to ameliorate perceived negative effects; nothing deliberate has to be done in the
case of transitory shocks since the economy may self-correct in due course. The Hodrick-
Prescott (HP) decomposition technique was employed. For the whole sample, and for 23 out
of the 24 countries, it was estimated that the shocks to government size were more of the
permanent type than the transitory type. It was for Australia only that the shocks were
deemed to be more of the transitory nature.
23
REFERENCES
Afonso, A. and Furceri, D. (2008) Government Size, Composition, Volatility and Economic
Growth, European Central Bank, Working Paper Series No. 849.
Afonso, A. and Jalles, J. T. (2011) Economic Performance and Government Size, European
Central Bank Working Paper Series, No. 1399, November 2011.
Agell, J., Lindh, T. and Ohlsson, H. (1997) Growth and Public Sector: A Critical Review
Essay, European Journal of Political Economy, Vol 13, No. 1, pp. 33-52.
Agell, J., Ohlsson, H. and Thoursie, P. S. (2006) Growth Effects of Government Expenditure
and Taxation in Rich Countries: A Comment, European Economic Review, Vol 50,
No. 1, pp. 211-219.
Armey, R. (1995) The Freedom Revolution, Rognery Publishing Inc., Washington, D.C.
Balatsky, E. (2012) Wagner’s Law, the Armey-Rahn Curve and the Paradox of Wealth,
Problems of Economic Transition, Vol 54, No. 12, pp. 40-54.
Banerji, A. and Dua, P. (2011) Predicting Recessions and Slowdowns: A Robust Approach,
Centre for Development Studies, Department of Economics, Delhi School of
Economics Working Paper No. 202.
Barro, R. (1990) Government Spending in a Simple Model of Endogenous Growth, Journal
of Political Economy, Vol 98, pp. S103-S125.
Barro, R. (1991) Economic Growth in a Cross Section of Countries, Quarterly Journal of
Economics, Vol 106, pp. 407-443.
Barro, R. and Sala-i-Martin, X. (1992) Convergence, Journal of Political Economy, Vol 100,
pp. 223-251.
Barro, R. and Sala-i-Martin, X. (2003) Economic Growth, 2nd
edn. Cambridge, MA: MIT
Press.
Baxter, M. and King, R. G. (1999) Measuring Business Cycles: Approximate Band-Pass
Filters for Economic Time Series, Review of Economics and Statistics, Vol 81, No. 4,
pp. 575-593.
Bergh, A. and Henrekson, M. (2011) Government Size and Growth: A Survey and
Interpretation of the Evidence, IFN Working Paper No. 858,
http://ssrn.com/abstract=1734206.
Beveridge, S. and Nelson, C. R. (1981) A New Approach to Decomposition of Economic
Time Series into Permanent and Transitory Components with Particular to
Measurement of the “Business Cycle”, Journal of Monetary Economics, Vol 7, pp.
151-174.
Boschan, C. and Ebanks, W. W. (1978) The Phase-Average Trend: A New Way of
Measuring Growth. In: 1978 Proceedings of the Business and Statistics Section,
American Statistical Association, Washington D.C.
Chen, S-T. and Lee, C. C. (2005) Government Size and Economic Growth in Taiwan: A
Threshold Regression Approach, Journal of Policy Modeling, Vol 27, pp. 1051-1066.
Chen, S-T., Chen, C-C. and Kim, Y. (2011) Economic Growth and Government Size in
OECD Countries: New Evidence from the Quantile Regression Approach, Economics
Bulletin, Vol 31, No. 1, pp. 416-425.
Chiou-Wei, S-Z., Zhu, Z. and Kuo, Y-H. (2010) Government Size and Economic Growth: An
Application of the Smooth Transition Regression Model, Applied Economics Letters,
Vol 17, pp. 1405-1415.
Chobawnov, D. and Mladenova, A. (2009) What is the Optimum Size of Government?,
Institute for Market Economics, Bulgaria, August, www.ime.bg.
24
Conte, M. A. and Darrat, A. F. (1988) Economic Growth and the Expanding Public Sector: A
Re-examination, The Review of Economics and Statistics, Vol 70, No. 2, pp. 322-330.
Dar, A. A. and AmilKhalkhali, S. (2002) Government Size, Factor Accumulation and
Economic Growth: Evidence from OECD Countries, Journal of Policy Modeling, Vol
24, pp. 679-692.
Davies, A. (2008) Human Development and the Optimal Size of Government, Journal of
Socioeconomics, forthcoming.
De Witte, K. and Moesen, W. (2010) Sizing the Government, Public Choice, Vol 145, pp.
39-55.
Duverall, D. and Henrekson, M. (2011) The Futile Quest for a Grand Explanation of Long-
Run Government Expenditure, Journal of Public Economics, Vol 95 Nos. 7 and 8, pp.
708-722.
Facchini, F. and Melki, M. (2011) Optimal Government Size and Economic Growth in
France (1871-2008): An Explanation by the State and Market Failures, CES (Centre
d’Economie de la Sorbonne) Working Papers, No. 2011.77
Feder, G. (1982) On Export and Economic Growth, Journal of Development Economics, Vol
12, pp. 59-73.
Fölster, S. and Henrekson, M. (2001) Growth Effects of Government Expenditure and
Taxation in Rich Countries, European Economic Review, Vol 45, pp. 1501-1520.
Forte, F. and Magazzino, C. (2010) Optimal Size of Government and Economic Growth in
EU-27, C. R. E. I. Working Papers No. 04, October. http://mpra.ub.uni-
muenchen.de/26669.
Garratt, A., Robertson, D. and Wright, S. (2006) Permanent vs Transitory Components and
Economic Fundamentals, Journal of Applied Econometrics, Vol 21, pp. 521-542.
Ghali, K. H. (1998) Government Size and Economic Growth: Evidence from a Multi-variate
Cointegration Analysis, Applied Economics, Vol 31, pp. 975-987.
Grimes, A. (2003) Economic Growth and the Size and Structure of Government: Implications
for New Zealand, Motu Working Papers, No. 03-10
Grossman, P. J. (1990) Government and Growth: Cross-section Evidence, Public Choice, Vol
65, pp. 217-227.
Guseh, J. S. (1997) Government Size and Economic Growth in Developing Countries: A
Political Economy Framework, Journal of Macroeconomics, Vol 19, pp. 175-192.
Gwartney, J. D., Lawson R. A. and Holcombe, R. G. (1998) The Size and Functions of
Government and Economic Growth, Paper prepared for the Joint Economic
Committee, Washington D.C.
Handoussa, H. and Reiffers, J. L. (2003) FEMISE Report on the Euro-Mediterranean
Partnership: Analysis and Proposals of the Euro-Mediterranean Forum of Economic
Institutes, Report presented at the Steering Committee of FEMISE, Marseilles,
France.
Herath, S. (2012) Size of Government and Economic Growth: A Nonlinear Analysis,
Economic Annals, Vol LVII, No. 194, pp. 7-30.
Illarionov, A. and Pivarova, N. (2002) Size of Government and Economic Growth, in
‘Voprosy Economiki’.
Koenker, R. (2005) Quantile Regression, Cambridge University Press.
Koenker, R. and Bassett, G. (1978) Regression Quantiles, Econometrica, Vol 46, pp. 33-50.
Kormendi, R. C. and Meguire, P. (1986) Government Department, Government Spending
and Private Sector Behavior, Reply, American Economic Review, Vol 76, No. 1, pp.
191-203.
25
Kuckuck, J. (2012) Testing Wagner’s Law at Different Stages of Economic Development – A
Historical Analysis of Five Western European Countries, Institute of Empirical
Economic Research, Working Paper 91, Osnabrueck University, Germany.
Laffer, A. (2004) The Laffer Curve: Past, Present and Future, Heritage Foundation
Backgrounder, No. 1765.
Lamartina, S. and Zaghini, A. (2012) Increasing Public Expenditure: Wagner’s Law in
OECD Countries, German Economic Review, Vol 12, No. 2, pp. 149-164.
Landau, D. (1985) Government Expenditure and Economic Growth in the Developed
Countries: 1952-76, Public Choice, Vol 47, pp. 459-477.
Myles, G. D. (2000) Taxation and Economic Growth, Fiscal Studies, Vol 21, No. 1, pp. 141-
168.
Murray, C. J. and Papanyan, S. (2004) The Relative Importance of Permanent and Transitory
Components in Macroeconomic Time Series, University of Houston, Manuscript, pp.
1-71.
New Zealand Treasury (2011) Government and Economic Growth: Does Size Matter?, New
Zealand Treasury Paper 11/01, April 2011.
New Zealand Treasury 2025 Taskforce (2010) Focusing on Growth: Closing the Income Gap
with Australia by 2025, Second Report – 2025 Taskforce, November 2010. URL:
http://www.2025taskforce.govt.nz/fromthetaskforce.htm.
Nijkamp, P. and Poot, J. (2004) Meta-analysis of the Effect of Fiscal Policies on Long-run
Growth, European Journal of Political Economy, Vol 20, pp. 91-124.
OECD (2011) ‘Final Consumption Expenditures by Government and Households’, in
Government at a Glance 2011, OECD Publishing.
http://dx.doi.org/10.1787/gov_glance-2011-16-en.
Padda, Ihtsham ul Haq (2011) Decomposition of Expenditure Rate Series into Permanent and
Transitory Parts, European Journal of Scientific Research, Vol 56 No. 1, pp. 29-43.
Peacock, A. T. and Wiseman, J. (1961) The Growth of Public Expenditure in the United
Kingdom, NBER Books, National Bureau of Economic Research, Inc.
Pevcin, P. (2004) Does Optimal Spending Size of Government Exist?, Paper presented at the
European Group of Public Administration Conference, 1-4 September 2004,
Ljubljana, Slovenia.
Rahn, R. and Fox, H. (1996) What is the Optimum Size of Government?, Vernon K. Krieble
Foundation.
Ram, R. (1986) Government Size and Economic Growth: A New Framework and some
Evidence from Cross-Section and Time-Series Data, American Economic Review, Vol
76, No. 1, pp. 191-203.
Radwan, S. and Reiffers, J. L. (2004) FEMISE Report on the Euro-Mediterranean
Partnership: Analysis and Proposals of the Euro-Mediterranean Forum of Economic
Institutes, Report presented at the Steering Committee of FEMISE, Marseilles,
France.
Saunders, P. (1985) Public Expenditure and Economic Performance in OECD Countries,
Journal of Public Policy, Vol 5, No. 1, pp. 1-21.
Saunders, P. (1985) Public Expenditures and Economic Performance in OECD Countries,
Journal of Public Policy, Vol 5, pp. 1-21.
Saunders, P. (1986) What can We Learn from International Comparisons of Public Sector
Size and Economic Performance?, European Sociological Review, Vol 2, No. 1, pp.
52-60.
Scully, G. (1998) Measuring the Burden of High Taxes, National Center for Policy Analysis,
Policy Report No. 215.
26
Scully, G. (2003) Optimal Taxation, Economic Growth and Income Inequality, Public
Choice, Vol 115, pp. 299-312.
Sheehey, E. (1993) The Effect of Government Size on Economic Growth, Eastern Economic
Journal, Vol 19, No. 3, pp. 321-328.
Singh, S. K. (2008) Public Finance in Theory and Practice, S. Chand, p. 35, ISBN 81-219-
1091-9.
Smith, D. (1975) Public Consumption and Economic Performance, National Westminster
Bank Quarterly Review, Nov., pp. 17-30.
Statistics New Zealand (2007) Extracting Growth Cycles from Productivity Indexes,
Statistics New Zealand, October 2007, Wellington.
Tanninen, H. (1999) Income Inequality, Government Expenditures and Growth, Applied
Economics, Vol 31, pp. 1109-1117.
Temple, J. (1999) The New Growth Evidence, Journal of Economic Literature, Vol 37, pp.
112-156.
Vedder, R. K. and Galloway, L. E. (1998) Government Size and Economic Growth, Paper
prepared for the Joint Economic Committee, Washington D.C.
Wagner, A. (1893) Grundlegung der Politischen Ӧkonomie, Teil I: Grundlagen der
Volkswirtschaft, C. F. Wintersche Verlagshandlung.
Wagner, A. (1911) ‘Staat (in nationalӧkonomischer Hinsicht)’, in Handwӧrterbuch der
Staatswissenschaften, pp. 727-739, Gustave Fisher Verlag.
Wooldridge, J. M. (2013) Introductory Econometrics: A Modern Approach, 5th
ed. South-
Western Cengage Learning, Mason, OH, USA.
Zarnowitz, V. and Ozyildirim, A. (2006) Time Series Decomposition and Measurement of
Business Cycles, Trends and Growth Cycles, Journal of Monetary Economics, Vol
53, pp. 1717-1739.
27
APPENDICES
Appendix 1: The Raw Data on Government Size
Year AUS AUT BEL CAN DNK FIN FRA GER GRC ISL IRL ITA JPN LUX NLD NZL NOR PRT ESP SWE CHE TUR GBR USA Avge
1973 14.4 15.2 18.5 20.0 21.6 15.6 17.8 18.5 10.2 15.3 17.4 16.6 11.9 11.5 20.6 15.0 17.5 11.2 10.2 23.0 8.6 9.1 18.8 16.8 15.6
1974 16.7 15.9 18.9 20.3 23.6 15.8 18.3 20.0 12.6 17.0 18.9 15.8 13.1 11.7 21.5 16.9 17.7 12.3 10.6 23.5 8.9 8.2 20.5 17.3 16.5
1975 17.8 17.4 21.1 21.8 24.8 17.5 19.8 21.1 13.7 17.4 20.6 16.2 14.4 15.3 22.9 18.1 18.7 13.1 11.2 24.1 9.7 9.2 22.4 17.9 17.8
1976 17.4 17.8 21.3 21.8 24.3 18.6 20.1 20.6 13.4 16.6 19.9 15.7 14.1 15.1 22.8 16.9 19.4 12.0 12.1 25.2 10.1 10.1 22.2 17.3 17.7
1977 17.8 17.6 21.8 22.3 24.2 19.1 20.4 20.5 14.2 16.7 19.0 16.0 14.1 16.3 23.5 18.1 19.6 12.2 12.3 27.9 10.0 11.2 20.8 17.0 18.0
1978 17.3 18.3 22.6 21.8 24.8 18.7 20.8 20.7 14.0 17.7 19.3 16.5 13.9 16.0 23.9 19.4 20.1 12.1 12.8 28.3 9.9 9.8 20.5 16.4 18.2
1979 17.1 18.1 22.9 21.1 25.5 18.3 20.8 20.6 14.4 18.3 20.2 16.6 13.9 16.4 24.7 19.2 19.6 12.1 13.3 28.7 9.9 9.3 20.2 16.2 18.2
1980 17.7 18.2 22.8 21.3 27.1 18.4 21.4 21.1 14.4 17.6 22.1 16.8 14.1 17.1 24.5 20.2 19.2 12.7 14.0 29.3 9.8 6.8 21.7 16.8 18.5
1981 17.6 18.6 24.0 21.2 27.9 19.0 22.2 21.6 15.8 17.9 22.1 18.2 14.1 17.8 24.5 19.8 19.4 13.1 14.9 29.6 9.9 8.8 22.3 16.6 19.0
1982 18.6 19.0 23.7 22.9 28.4 19.2 22.7 21.3 16.0 18.9 21.8 18.3 14.2 16.8 25.0 19.7 19.7 13.0 15.1 29.5 10.2 7.6 22.2 17.6 19.2
1983 18.2 19.0 23.4 22.7 27.6 19.6 22.8 21.0 16.6 18.9 21.4 18.7 14.5 16.1 24.9 18.8 19.7 13.2 15.6 28.8 10.5 8.4 22.0 17.5 19.2
1984 18.8 19.0 23.3 21.8 25.7 19.6 23.0 20.8 17.1 17.4 20.9 18.5 14.3 15.7 23.4 18.1 18.8 13.1 15.3 27.8 10.3 7.2 21.7 17.1 18.7
1985 18.9 19.2 22.8 21.8 25.3 20.6 23.1 20.8 17.8 18.3 20.8 18.6 13.9 16.1 23.7 18.3 18.5 13.5 15.6 27.5 10.4 7.2 20.9 17.5 18.8
1986 18.9 19.5 22.7 21.7 24.2 20.8 22.8 20.6 16.8 18.8 21.1 18.3 13.9 15.8 23.7 18.5 19.6 13.4 15.4 27.0 10.5 7.3 20.9 17.8 18.7
1987 17.9 19.4 22.4 21.1 25.3 21.2 22.6 20.9 17.2 19.6 20.1 19.0 13.9 16.7 24.6 18.1 20.7 13.2 15.9 26.2 10.5 6.3 20.3 17.8 18.8
1988 17.3 19.2 21.0 20.9 25.8 20.5 22.2 20.6 15.1 20.7 18.7 19.4 13.5 15.8 24.0 18.0 20.9 13.7 15.7 25.6 10.7 6.1 19.6 17.3 18.4
1989 17.1 18.8 19.9 21.1 25.5 20.2 21.7 19.6 16.0 20.1 17.5 19.2 13.4 15.4 23.2 18.1 20.7 14.4 16.3 25.8 11.1 7.5 19.4 16.9 18.3
1990 18.0 18.6 19.7 22.3 25.1 21.8 21.7 19.3 16.1 19.9 17.9 20.0 13.3 15.8 23.0 18.8 21.2 15.2 16.7 26.9 11.3 8.8 19.7 17.0 18.7
1991 18.9 18.8 20.6 23.7 25.3 24.9 22.2 19.1 15.2 20.6 18.9 20.1 13.4 15.3 23.2 19.2 21.8 16.9 17.4 27.6 11.7 10.0 20.7 17.2 19.3
1992 18.8 19.1 20.9 24.1 25.3 25.4 22.8 19.6 14.7 21.2 19.3 20.0 13.8 15.9 23.8 19.3 22.7 16.9 18.3 28.7 12.2 10.4 21.2 16.8 19.6
1993 18.2 20.0 21.2 23.5 26.4 24.2 24.0 19.6 15.3 21.7 19.1 19.9 14.3 15.8 24.1 18.1 22.5 17.5 18.8 28.8 12.0 10.5 20.4 16.2 19.7
1994 17.9 20.1 21.2 22.3 25.5 23.5 23.7 19.5 14.7 21.6 18.9 19.2 14.7 15.4 23.9 17.2 22.2 17.6 18.2 27.9 11.9 9.4 20.0 15.7 19.2
1995 17.8 20.3 21.4 21.3 25.2 22.7 23.6 19.4 16.4 22.1 17.7 17.8 15.2 15.9 23.8 17.2 21.6 17.5 18.1 26.6 11.8 8.7 19.5 15.4 19.0
1996 17.5 20.2 21.8 20.5 25.4 23.2 23.9 19.7 15.5 21.9 17.1 18.1 15.4 16.4 22.8 17.0 20.9 17.7 18.0 27.3 11.9 9.3 19.0 15.0 19.0
28
1997 17.5 19.3 21.3 19.5 25.0 22.5 23.9 19.3 16.2 21.7 16.3 18.2 15.4 16.7 22.3 17.8 20.6 17.7 17.5 26.7 11.6 9.9 18.0 14.6 18.7
1998 17.8 19.2 21.2 19.6 25.6 21.8 23.1 19.1 16.4 22.2 15.5 18.0 15.9 15.8 22.2 17.7 22.0 17.9 17.3 26.8 11.3 10.3 17.8 14.3 18.7
1999 17.6 19.6 21.4 18.9 25.7 21.4 23.2 19.2 16.5 22.9 14.9 18.1 16.5 15.5 22.2 18.0 21.6 18.1 17.2 26.7 11.1 12.2 18.3 14.3 18.8
2000 17.6 19.0 21.3 18.6 25.1 20.6 22.9 19.0 18.9 23.4 14.7 18.3 16.9 15.1 22.0 17.2 19.3 19.0 17.1 25.8 11.1 11.7 18.7 14.3 18.7
2001 17.4 18.7 21.7 19.1 25.7 20.7 22.8 19.0 18.4 23.6 15.5 18.8 17.7 16.1 22.6 17.2 20.6 19.4 17.0 26.3 11.6 12.4 19.1 14.8 19.0
2002 17.5 18.4 22.5 19.5 26.2 21.4 23.5 19.2 19.4 25.4 16.0 19.0 18.3 16.5 23.7 17.0 22.1 19.7 17.1 27.0 11.8 12.7 19.9 15.4 19.6
2003 17.4 18.7 22.9 19.7 26.5 22.1 23.8 19.3 18.1 26.0 16.1 19.5 18.3 16.4 24.5 17.2 22.5 20.0 17.3 27.3 12.0 12.2 20.5 15.8 19.8
2004 17.4 18.4 22.5 19.2 26.5 22.2 23.8 18.9 18.3 25.0 16.4 19.7 18.2 16.9 24.2 17.5 21.2 20.3 17.8 26.5 11.8 11.9 20.9 15.8 19.6
2005 17.3 18.4 22.7 18.9 26.0 22.5 23.8 18.8 18.1 24.6 16.3 20.1 18.4 16.5 23.7 18.0 19.7 21.1 18.0 26.2 11.6 11.8 21.2 15.8 19.6
2006 17.2 18.3 22.4 19.1 25.9 22.2 23.5 18.4 17.0 24.4 16.5 20.0 18.2 15.4 25.1 18.5 18.9 20.5 18.0 26.0 11.1 12.3 21.4 15.8 19.4
2007 17.1 18.0 22.2 19.2 26.0 21.5 23.1 17.9 17.8 24.2 17.2 19.5 18.1 14.8 25.2 18.6 19.3 19.8 18.3 25.5 10.7 12.8 20.9 15.9 19.3
2008 17.6 18.7 23.1 19.7 26.5 22.5 23.3 18.3 18.1 24.8 19.2 20.0 18.6 14.8 25.7 20.2 19.1 20.1 19.5 26.1 10.4 12.8 21.9 16.9 19.9
2009 18.1 19.8 24.7 22.1 29.8 25.2 24.8 20.0 20.4 26.5 20.4 21.4 19.9 16.9 28.6 20.3 22.5 22.1 21.3 27.7 11.2 14.7 23.4 17.9 21.6
2010 17.9 19.4 24.3 21.8 29.1 24.7 24.9 19.5 18.2 25.9 19.2 21.1 19.8 16.6 28.4 20.1 22.0 21.6 21.4 26.7 11.0 14.3 22.8 17.9 21.2
2011 17.7 18.8 24.4 21.4 28.6 24.3 24.5 19.3 17.5 25.3 18.4 20.5 20.6 16.5 27.9 20.2 21.5 20.1 20.9 26.4 11.1 13.9 22.4 17.3 20.8
Year AUS AUT BEL CAN DNK FIN FRA GER GRC ISL IRL ITA JPN LUX NLD NZL NOR PRT ESP SWE CHE TUR GBR USA Avge
Avge 17.7 18.7 22.0 21.0 25.9 21.1 22.5 19.8 16.2 21.1 18.5 18.6 15.5 15.8 24.0 18.3 20.4 16.3 16.3 26.9 10.9 10.1 20.6 16.5 18.9
Sources: World Bank and OECD databases.
29
Appendix 2
Country-wise Correlations of Government Size
AUS AUT BEL CAN DNK FIN FRA GER GRC ISL IRL ITA JPN LUX NLD NZL NOR PRT ESP SWE CHE TUR GBR USA
AUS 1.00
AUT 0.66 1.00
BEL 0.42 0.40 1.00
CAN 0.55 0.20 0.06 1.00
DNK 0.40 0.48 0.80 0.15 1.00
FIN 0.47 0.72 0.29 0.15 0.49 1.00
FRA 0.49 0.82 0.60 -0.09 0.68 0.84 1.00
GER 0.37 0.06 0.28 0.54 0.13 -0.43 -0.28 1.00
GRC 0.40 0.55 0.61 -0.32 0.62 0.58 0.81 -0.27 1.00
ISL 0.08 0.44 0.37 -0.42 0.54 0.75 0.80 -0.65 0.79 1.00
IRL 0.39 -0.01 0.36 0.71 0.24 -0.25 -0.22 0.79 -0.27 -0.56 1.00
ITA 0.33 0.52 0.36 0.04 0.61 0.83 0.78 -0.43 0.69 0.77 -0.18 1.00
JPN 0.01 0.24 0.57 -0.47 0.61 0.54 0.68 -0.52 0.76 0.90 -0.42 0.59 1.00
LUX 0.55 0.64 0.71 0.15 0.66 0.43 0.63 0.34 0.53 0.31 0.19 0.34 0.31 1.00
NRL 0.31 0.35 0.80 0.24 0.84 0.50 0.56 0.07 0.51 0.50 0.35 0.60 0.61 0.50 1.00
NZL 0.51 0.29 0.65 0.40 0.71 0.32 0.33 0.28 0.29 0.16 0.53 0.39 0.25 0.52 0.75 1.00
NOR 0.35 0.66 0.14 0.22 0.42 0.79 0.63 -0.22 0.38 0.60 -0.28 0.58 0.34 0.46 0.32 0.15 1.00
PRT 0.07 0.40 0.30 -0.41 0.48 0.77 0.77 -0.72 0.73 0.97 -0.59 0.74 0.88 0.24 0.42 0.11 0.55 1.00
ESP 0.36 0.71 0.44 -0.07 0.65 0.92 0.93 -0.49 0.75 0.87 -0.30 0.89 0.72 0.44 0.60 0.37 0.68 0.86 1.00
SWE 0.61 0.54 0.55 0.49 0.52 0.29 0.40 0.44 0.19 -0.02 0.43 0.18 -0.06 0.72 0.35 0.56 0.36 -0.05 0.24 1.00
CHE 0.38 0.71 0.09 -0.03 0.30 0.85 0.78 -0.45 0.56 0.72 -0.50 0.66 0.44 0.45 0.16 -0.02 0.82 0.72 0.78 0.27 1.00
TUR -0.13 0.03 0.32 -0.34 0.39 0.50 0.44 -0.63 0.50 0.76 -0.49 0.45 0.85 0.09 0.43 0.15 0.36 0.81 0.57 -0.10 0.38 1.00
GBR 0.33 -0.06 0.63 0.44 0.57 0.06 0.09 0.38 0.15 0.01 0.69 0.21 0.25 0.25 0.71 0.63 -0.13 0.00 0.09 0.24 -0.24 0.17 1.00
USA 0.24 -0.21 0.21 0.64 0.17 -0.16 -0.27 0.53 -0.19 -0.38 0.80 0.04 -0.25 -0.05 0.42 0.43 -0.24 -0.43 -0.22 0.02 -0.48 -0.30 0.73 1.00
30